Climate Reconstructions Derived from Global Glacier Length Records

Arctic, Antarctic, and Alpine Research, Vol. 36, No. 4, 2004, pp. 575-583

E. J. Klok* t and Abstract J. Oerlemans* As glacier length fluctuations provide useful information about past climate, we derived *Institute for Marine and Atmospheric historic fluctuations in the equilibrium-line altitude (ELA) on the basis of 19 glacier length Research Utrecht, Utrecht University, records from different parts of the world. We used a model that takes into account the Princetonplein 5, 3584 CC Utrecht, geometry of the glacier, the length response time and the mass balance-surface height The Netherlands feedback. The results show that all glaciers of the data set experienced an increase in the tE.J.Klok@ phys.uu.nl ELA between 1900 and 1960. Between 1910 and 1959, the average increase was 33 ± 8 m. This implies that during the first half of the 20th century, the climate was warmer or drier than before. The ELAs decreased to lower elevations after around 1960 up to 1980, when most of the ELA reconstructions end. These results can be translated into an average temperature increase of 0.8 ± 0.2 K and a global sea-level rise of about 0.3 mm a-1 for the period 19/0-/959.

Introduction Oerlemans (1994) compared the retreat of 48 glaciers from different regions. He estimated from this a global linear warming trend of 0.66 K Glacier length records contain information on how climate has per century, but did not take the response time of the glaciers into changed. This information often complements historical meteorolog- account. Hoelzle et al. (2003) also investigated 90 glaciers worldwide. ical data, as glacier length records generally extend farther back in They estimated, from cumulative glacier length changes, a global mean time. Besides, glacier records are often from remote areas and higher specific mass balance of -0.25 m w.e. a l since 1900. Furthermore, the altitudes, for which meteorological data are scarce (Intergovernmental IPCC (2001) analyzed a collection of glacier length records from Panel on Climate Change [IPCC], 2001). Hence, glacier length records different parts of the world quantitatively. They argued that glacier form an alternative method for climate reconstruction for periods and retreat already began around 1850 in the low and mid-latitudes, but locations for which instrumental or proxy indicators are inadequate or started later at high latitudes. contradicting. This studythusfocuses on theclimaticinterpretationof From many glaciers in Europe, we know how they have retreated or worldwide glacier length changes and uses a more sophisticated model advanced since 1850 or even earlier. In an earlier study (Klok and than those used by Oerlemans (1994) and Hoelzle et al. (2003), Oerlemans, 2003), we used 17 European length records to retrieve because we correct for the length response time and treat the geometry climate reconstructions. The reconstructions were described in terms of of each glacier more comprehensively. Building a data set of useful changes in the equilibrium-line altitude (ELA). ELA fluctuations length records was difficult because most non-European glaciers have resemble transient climate changes more properly than glacier length only been measured since the beginning of the 20th century. The first changes, as the glacier length is subject to the length response time and criterion for populating our data set was to only use records that the sensitivity of a glacier. The model that we developed for extracting predate AD 1900. A second screening criterion of our data set was that ELA reconstructions from glacier length fluctuationsisa simple the length fluctuations must be small (<--20%) compared to the total analytical one, based on the assumption that the change in length can be glacier length, to justify the assumption of a quasi-linear response. described by a linear response equation. Its advantage is that it takes the Third, only valley or outlet glaciers are included. Lastly, we also length response time and the geometry of the glacier as well as the mass needed information about the geometry and the mass balance to balance-surface height feedback into account. The length response time calculate the response time and the climate sensitivity. This led to expresses the speed of the glacier response to a change in climate or a collection of 19 glaciers: 14 from countries other than European, and mass balance. Results from this model indicate that the ELA of 17 5 European yielding a global sample (Fig. 1). European glaciers increased on average 54 m between 1920 and 1950 In the following section, we describe the glaciers of our data set in (Klok and Oerlemans, 2003). ELA reconstructions modeled with this more detail. To these glaciers, we applied the analytical model as analytical model agree well with mass-balance reconstruction derived explained in Klok and Oerlemans (2003), without any modifications, from numerical flowline models and from historical temperature and and we only explain the model briefly here. As we did not have precipitation records (Klok and Oerlemans, 2003). In this study, we accurate data on the glacier geometry or the mass-balance regime for extend this investigation to glaciers in other parts of the world. some glaciers, we investigated the effects of the uncertainties in the So far, the climatic interpretation of worldwide glacier length input data on the ELA reconstruction. The results of the ELA and fluctuations has been studied occasionally. More frequently, length mass-balance reconstructions are also described and interpreted in fluctuations of single glaciers or glaciers in one specific region were terms of changes in air temperature. investigated, e.g., the European Alps (Haeberli and Hoelzle, 1995), the Central Italian Alps (Pelfini and Smiraglia, 1997), Scandinavia (Bogen et al., 1989), the Patagonian Ice Fields (Warren and Sugden, 1993; Glaciers and Their Length Records Aniya, 1999), the North Cascade glaciers (Pelto and Hedlund, 2001), Northern Eurasia (Solomina, 2000), Tien Shan (Savoskul, 1997), New The glaciers of our data set are located in six regions: Canada, Zealand (Chinn, 1996), and the tropics (Kaser,1999). However, U.S.A., South America, Europe, Asia, and New Zealand. All of them

..- ,

Clend 1 Havoc

Athab i - ...... White Blue

South Frias Sanqu

SOrbr

Nigar Argen 1

Morte Paste

Maruk

Bezen Gerge

Softy

Fox

Franz FIGURE 1.Glacier length fluctuations of the 19 glaciers of our data set. The first five letters of 1600 1650 1700 1750 1800 1850 1900 1950 2000 their names abbreviate the glacier year names. have retreated over the previous century, but 60% of them also slightly Mountains is strongly maritime and receives the greatest precipitation advanced after 1950 (Fig. 1). Retreat rates, total retreat and periods of of any area in the U.S.A., excluding Alaska and Hawaii (Armstrong,

retreat vary widely between the glaciers. Table I gives information 1989). Blue Glacier is steeper and narrower than White Glacier and has about the geometry and the mass balance of the glaciers. These retreated less. South Cascade is in the North Cascade Range, 250 km parameter values were used as model input. The length records and the from the ocean. This glacier receives less snow (--3 m w.e.) than White input data were either collected or estimated from data of the World and Blue Glaciers (--4 m w.e.) (Rasmussen and Conway, 2001). Mass- Glacier Monitoring Service,the World Glacier Inventory, from balance measurements have been made on Blue Glacier (Conway et al., Dyurgerov (2002) or from other references mentioned in the text. 1999) and South Cascade (Krimmel, 2000). Armstrong (1989) showed that Blue Glacier had a slightly positive mass balance between 1956 CANADA and 1986 (0.3 m) and Krimmel (1989) found a mean negative balance for South Cascade (-0.22 m) for 1959 to 1985. Blue Glacier has not The three Canadian glaciers of the data set are Athabasca retreated significantly since 1950, while the length record of South (Rockies),Clendenning(CoastMountains),and Havoc(Coast Cascade shows an ongoing retreat. Rasmussen and Conway (2001) Mountains). Athabasca is an outlet glacier of the Columbia Ice Field, concluded that South Cascade has been more out of balance than White an ice cap of roughly 325 km2. The glacier flows over three icefalls Glacier because of differences in geometry between the two glaciers. into an alpine valley (Reynolds and Young, 1997). It has retreated rapidly since 1910 until 1960, after which a slowdown in the retreat rate occurred. Clendenning and Havoc, two valley type glaciers, are SOUTH AMERICA located closer to the ocean and their glacier termini reach a lower elevation (- 1000 m a.s.l.) than Athabasca (- 1950 m a.s.l.). The length Frias Glacier (Argentina) is an outlet glacier on the east side of records of Clendenning and Havoc do not show similar patterns, Mount Tronador. This ice cap isthe northernmost ice body in although both are located in the Clendenning valley. This could be Argentina. Measurements from a weather station in the Rio Frias attributed to differences in glacier geometry: Havoc is much steeper Valley of Frias Glacier indicated that the annual precipitation amounts and narrower. to 4300 mm a-' (Perez Moreau, 1945). The glacier fluctuations have been recorded since 1976 and Villalba et al. (1990) dated oscillations of Frias Glacier using tree rings. Between 1850 and 1900, the retreat U.S.A. rate was 7 mat and increased to 10 mat between 1910 and 1940. White and Blue Glaciers are located in the Olympic Mountains, San Quintin is farther south than Frias Glacier and is the largest glacier about 55 km from the Pacific Ocean. The climate of the Olympic of our data set. It is a piedmont outlet glacier of the North Patagonian

576 / ARCTIC, ANTARCTIC, AND ALPINE RESEARCH TABLE I Model parameters of the 19 glaciers, the calculated length response time (t,.L), climate sensitivity (c), an the mean annual precipitation (Zuo and Oerlemans, 1997b). The first five letters of their names abbreviate the glacier names. (Lo: reference glacier length, A0: reference glacier area, s: mean surface slope, sf: surface slope of the glacier snout, wf characteristic width of the glacier snout, Bf> melt rate at the glacier terminus, (3E: mass balance gradient at the ELA, (3: mass balance gradient of the glacier)

Glacier L0 (km) A0 (km) s (%) sf (%) wf (m) Bf(m w.e.a-') a (ie nc (a) c P,,.,,, (m a-')

Clend 11.1 26.5 15 8 500 -8 0.0052 0.0065 80 49 2.5 Havoc 7.0 9.5 19 30 250 -6.9 0.0052 0.0065 46 36 2.5 Athab 9.3 15.0 17 8 550 -4.7 0.0058 0.0064 102 43 0.75 White 3.3 3.8 27 35 290 -5.6 0.010 0.010 18 16 4 Blue 4.3 5.5 27 160 400 -5.7 0.010 0.010 8 10 4 South 3.1 2.9 16 10 400 -6 0.0105 0.023 27 26 2.5 Frias 6 10.5 21 10 400 -12 0.0110 0.017 33 38 3 Sanqu 60 765.0 3 5 8500 -15.8 0.0150 0.015 76 128 2 SOrbr 9 15 20 8 800 -4 0.005 0.006 105 32 0.7 Nigar 9.6 48.2 17 23 500 -10 0.0079 0.0086 68 84 4 Argen 9.4 15.6 16 16 400 -12.2 0.0078 0.0071 34 25 2 Morte 7.0 17.2 29 6 650 -6.5 0.0046 0.0065 88 29 2 Paste 9.4 19.8 17 10 750 -5.8 0.004 0.007 62 32 2 Maruk 4.0 3.1 17 10 500 -2.5 0.0100 0.0061 91 19 1.5 Bezen 17.6 31.1 17 10 450 -7.2 0.0014 0.0042 85 42 1.5 Gerge 8.5 7.1 30 20 200 -3.9 0.0014 0.0042 78 37 1.5 Softy 7.0 20 20 10 500 -5 0.0054 0.0073 97 63 0.375 Fox 13.2 34.7 24 5 350 -23 0.011 0.017 56 92 1.87 Franz 10.3 32.6 25 8 425 -22 0.011 0.017 38 72 1.87

Ice Field in Chile. San Quintin flows to the west onto an outwash plain. measurements, which explains the straight lines in Figure1. The Itreceives between 3700 and 6700 mm precipitation per year. mass-balance gradient of these glaciers is relatively small and the melt Winchester and Harrison (1996) investigated fluctuations of the ice rate at the glacier terminus is low (Table 1). The maximum elevation of front and concluded that precipitation is the main factor controlling the Bezengi and Gergeti is over 5000 m a.s.l. Sofiyskiy glacier is located terminus position. The length record of San Quintin shows a peculiar in the Altai Mountains of central Asia, a border region between glacier retreat just before 1935 from which it recovers after 1935, Russia andMongolia.Itisaso-calledcontinental,summer- which does not show up in the other length records. accumulation type glacier. Sofiyskiy glacier is a valley type glacier consisting of three basins. The most rapid retreat of Sofiyskiy glacier occurred between 1900 and 1940 (De Smedt and Pattyn, 2003; Pattyn EUROPE et al., 2003). Sorbreen is a glacier on Jan Mayen, the northernmost island on the Mid-Atlantic Ridge (71°N, 8°W). This glacier flows southwards to the ocean, from the 2277-m high Beerenberg volcano. The climate is NEW ZEALAND cool oceanic with a mean annual air temperature of -1.2°C at sea level. Fox and Franz Josef are valley glaciers in the maritime climate of The island is surrounded with pack ice during winter and spring (Andy the Southern Alps of New Zealand. They receive between 5 and 15 in et al., 1985). Nigardsbreen is an outlet glacier in Norway, flowing from precipitation per year. The glaciers reach very low elevations: 425 and thelargesticecapof continentalEurope (Jostedalsbreen)in 305 m a.s.l. for Franz Josef and Fox Glaciers, respectively (Hooker and a southeasterly direction. It is a maritime glacier, located close to the Fitzharris, 1999). The length records show similar patterns: a rapid Atlantic Ocean. It advanced very rapidly between 1710 and 1748 and retreat after 1940 and a major advance after 1982. The advance phase is after 1988, it slightly advanced again. Pohjola and Rogers (1997) characterized by higher precipitation and is also related to a higher claimed that this advance was due to enhanced westerly maritime flow frequency of El Nino events (Hooker and Fitzharris, 1999). after 1980 leading to high winter accumulation and low summer ablation. Glacier d'Argenti6re (France), Morteratschgletscher (Switzer- land), and Pasterzenkees (Austria) are valley glaciers in the European Alps, where the climate is more continental. Glacier d'Argentiere has been well documented and numerous fluctuations were registered Explanation of the Model before 1850. Its length record is the longest of our data set and has been A comprehensive description of the model that we used is given studied by Huybrechts et al. (1989). Glacier d'Argentiere advanced in Klok and Oerlemans (2003) and Oerlemans (2001). In this section, after 1970, in contrast to Morteratschgletscher and Pasterzenkees. we explain only the important aspects. The model is based on the Pasterzenkees isthe longest glacier of the Austrian Alps.It has following equation that relates changes in the ELA to glacier length experienced two periods of retreat: between 1870 and 1910 and after fluctuations: 1930 (Zuo and Oerlemans, 1997a).

(1) ASIA E'(t) is the ELA with regard to a reference altitude, L'(t) is the glacier Marukskiy, Bezengi, and Gergeti are valley glaciersinthe length with regard to a reference length (LO), t is time, t,.L is the length Caucasus Mountains. Their length records contain only a few response time, and c the climate sensitivity of the glacier. The latter

E. J. KLOK AND J. OERLEMANS / 577 TABLE 2 Parameters of the fictitious glacier

Parameter Value

L0 15km An 20 km2

S 10°

Sf to° wf 500 m Bf -5 m w.e.a-'

0 0.005 m w.e.a-' m-1 OF 0.007 m w.e.a-' m-'

I I I -20 -10 0 10 20 30 relates the change in the glacier's steady-state length to a change in the Change in parameter value (%) ELA. We calculated trL and c from: r1A0 + wfHf trL =- nlA0 + wfBf

READ C = (3) i1(3Ao + wfBf rl is a constant that relates the change in glacier thickness to a change in glacier length. We derived this parameter from the mean glacier slope (s) and LO. A0 is the reference area, wf is the characteristic width of the glacier snout, and Hf the characteristic thickness of the glacier snout, which we estimated from s and the surface slope of the glacier snout (sf). Q is the average mass-balance gradient over the glacier, 1'E the mass-balance gradient at the equilibrium-line altitude, and Bf the I I I I I -20 -10 0 10 20 30 melt rate at the glacier terminus. Change in parameter value (%) Since the length records contain data gaps and a continuous length -0 Lo n s -a - 0 record is needed for application of equation (1), we applied linear r interpolation between the data points of the length records. The --- 0- A-0 w T P f E reference length (Lo) was then defined as the mean length of this LS-X B interpolated record. We applied a Gaussian filter to the records and f calculated time derivatives with central differences to obtain a smooth FIGURE 2.Length response time (A) and climate sensitivity (B) as seriesof timederivatives(dL'(t)/dt).As we concentrated on function of a change in the input parameters. In B, Bfand wfcoincide. fluctuations in the ELA occurring on a decadal timescale, the timescale (LO: reference glacier length, A0: reference glacier area, s: mean of the Gaussian filter was taken as 10 yr. surface slope, sf surface slope of the glacier snout, wf characteristic This model thus takes into account the length response time of the width of the glacier snout, Bf; melt rate at the glacier terminus, (3E: glacier. This implies that the ELA reconstruction shifts backward in time mass balance gradient at the ELA, (3: mass balance gradient of the as the length response time increases and also influences the amplitude of glacier). the ELA fluctuations. The geometry of the glacier is included in the model, mainly defined by the characteristic width of the glacier snout and rate at the glacier terminus, the width of the glacier snout and the mass- the glacier area. Glaciers with relatively narrow tongues compared to the balance gradient at the equilibrium-line altitude. Also, the glacier area total surface area will have larger climate sensitivities and response times seems to influence the value of t,.L and c substantially. Nevertheless, we (eqs. 2, 3). Furthermore, the effect of the mass balance-surface height do not expect the surface area to be an important contributor to the feedback is included, by the integration of rl. For steep glaciers, rl is uncertainties in trL and c, as it is often known within 10%. An increase small (the surface height changes little when the glacier length changes), in the melt rate at the glacier terminus or an increase in its width both implying a weak mass balance-surface height feedback, resulting in lead to smaller response times, as both of them cause a larger mass smaller length response times and lower climate sensitivities. turnover. However, an increase in the characteristic width of the glacier snout also implies that more ice needs to be transported down the Sensitivities glacier to make up for an equal change in glacier length. This counterbalances the effect of a larger mass turnover on trL. Therefore, Climate sensitivity and length response time are determined by the length response time is more sensitive to melt rate than to width of eight parameters (Table 1). The values for these parameters in Table 1 the glacier snout. Regarding climate sensitivity, changes inthe may contain errors because some of them were unknown or difficult to characteristic width of the glacier snout or melt rate at the glacier estimate. We therefore investigated the effect of a change in these terminus lead to the same effect. Generally, trL and c vary in the same parameters on length response time and climate sensitivity and, finally, direction when one of the input parameters is varied. on the ELA reconstruction of a fictitious glacier. Table 2 lists the As a second step, we determined the effect of a change in length parameter values of this fictitious glacier. The values represent an response time and climate sensitivity on the ELA reconstruction of this average valley glacier and they yield a length response time of 85 yr fictitious glacier. We derived historic ELAs from a (fictitious) length and a climate sensitivity of 70. We changed each of these parameters record, which we defined as a sine function with an amplitude (L;nax) by 30%. The resulting length response times and climate sensitivities of 1.5 km and a period (P) of 300 yr. Instead of smoothing this record are plotted in Figure 2. They are most sensitive to changes in the melt with a Gaussian filter and calculating the derivatives with central

578 / ARCTIC, ANTARCTIC, AND ALPINE RESEARCH 2 60 60 m

1.5 45 55 d 1 30 8 50

0.5 15M E 45 0 t 03 CL 40 -0.5 -15 E 35 -1 -30

E 30 -1.5 -45 70 80 90 100 110 120 Length response time (a) -2 t I I I I -60 0 50 100 150 200 250 300 time (a) u60 55 FIGURE 3.Length record and reconstructed ELAs of the fictitious ar glacier. The length response time is 85 yr and the climate sensitivity 70. 450

u) 45 rS. differences, we directly inserted this function and its time derivative in 40 equation (1). This leads to: E '035 E'(t) _ -7 sin(21rtt (4) E 30 60 70 80 90 100 where 7. and (p are the amplitude and the phase difference of the ELA Climate sensitivity reconstruction (see Fig. 3 for explanation of the symbols) described by: 60

2 - 50 'k= 1 + (2rz P) Lmx (5) ar m40 1 P / P (p=4P-arctan E 2n 2ntrL) ¢ 30 -20 -10 0 10 20 Figure 3 showsthe length and the reconstructed ELA asfunctionof Percentage change inBt time. The amplitude of the ELA reconstruction thus depends on the period of the length record, the length response time and the climate FIGURE 4.Amplitude (solid) and phase (dashed) of the ELA sensitivity. The reconstruction experiences a phase difference depend- reconstruction of the fictitious glacier, when trL is varied and c kept at ing on the length response time. For very large length response times, 70 (A), when c is varied and trL is kept constant at 85 a (B), and when the phase difference approaches 0.25P. Bf is varied by 25% (C). We varied the length response time between 65 and 123 yr and the climate sensitivity between 53 and 101. These numbers are the limits set by a change of 25% (plus or minus) in the most sensitive parameter, indicates that the uncertainty in the characteristic width of the glacier melt rate at the glacier terminus (Fig. 2). The effects on the amplitude snouthasthelargestimpactonthe ELA reconstructionof and phase difference of the ELA reconstruction are plotted in Figure 4. Nigardsbreen, leading to a standard deviation of 4 m in E'(t), and The amplitude varies roughly from 30 to 59 m and the phase difference a maximum deviation in the ELA of 9 m for the maximum ELA, which from 18 to 30 yr. Note that an increase in t,.L leads to larger amplitudes occurred in 1978. and an increase in c to lower amplitudes. The total effect on the ELA reconstruction constitutes the sum of both, and is therefore smaller than the individual effects. Figure 4C shows that the total effect on the amplitude of the ELA reconstruction is at maximum 10%, when the melt rate at the glacier terminus is varied by 25%. On the basis of this sensitivity test, we conclude that the length response time and the climate sensitivity are most sensitive to uncertainties in the characteristic width of the glacier snout, the melt rate at the glacier terminus and the mass-balance gradient at the equilibrium- line altitude. The effect of these uncertainties on the final ELA reconstruction is, at most, 10% in the amplitude, and a phase difference of 6. Of course, this test provides no evidence that an uncertainty in the input parameters alwaysleadstosmallchangesinthe ELA reconstructions because other glaciers may have differentinput parameters and length records than our fictitious glacier. Therefore, we also carried out a sensitivity test for Nigardsbreen, a glacier for 1800 1850 1900 which we had a long length record. We investigated the effect on the Year ELA reconstruction when the most sensitive parameters (Bf, wfand Re) FIGURE 5.ELA reconstruction for Nigardsbreen for different values were changed within their uncertainty ranges (Fig.5). This test of the input parameters.

E. J. KLOK AND J. OERLEMANS / 579 140 (see also Table I for the precise values of trL and c). Maritime glaciers Sanqu often have high climate sensitivities (San Quintin, Fox, Nigardsbreen, 120 and Franz Josef), since they have large mass-balance gradients.' However, the glaciers of the U.S.A. (South Cascade, White, and Blue), 100 which are also located in maritime areas, have low climate sensitivities. .)Z. -80 Nigar This is due to their small surface areas and relative wide glacier tongues. Franz Athabasca Glacier has the largest length response time, which is partly y Softy 260 caused by its low melt rate at the glacier terminus. Blue Glacier has the M Clend smallest length response time and the lowest climate sensitivity. Since E Athab 040 Frtas Hivoc Gerge 6ezen this glacier is situated on a steep slope, the contribution of the mass Mortt balance-surface height feedback to tL and c is small. The calculated South Paste Mort% Serbr 20 rgen length response times and climate sensitivities generally agree well with Maruk values calculated from numerical flowline models, as was shown for BlueWhite _j I 0 European glaciers in Klok and Oerlemans (2003). 0 20 40 60 80 100 120 Length response time (a)

FIGURE 6.Climate sensitivity and length response time for all glaciers. ELA RECONSTRUCTIONS The first five letters of their names abbreviate the names of the glaciers. Figure 7 shows the reconstructed ELAs. Athabasca and Havoc show similar fluctuations in the ELA although the ELA of Havoc started to increase some years before 1900 and that of Athabasca some Results years after 1900. The ELA reconstruction of Clendenning shows neither any resemblances with the other Canadian ELAs nor with the LENGTH RESPONSE TIME AND CLIMATE SENSITIVITY reconstructions for the U.S.A. because the ELA does not stabilize We calculated the length response time (eq. 2) and climate around 1950. The reconstructed ELAs of White and Blue Glaciers are sensitivity (eq. 3) for all 19 glaciers and plotted the values in Figure 6 similar, but the fluctuation in White Glacier's ELA is larger.

120,

-30 A 18 00 1850 1900 1950 2000 Year

120 120 South America Europe 90 ---Frfas Serbreen -- San Quintin 90 ------Nigardsbreen Argentiere -S - Morteratsch 60 - - Pasterze w30 q,

-301-

Ic 1- r r 1 1 -80 r -601 - 1800 1850 1900 1950 2000 1800 1850 1900 1950 2000 Year Year

120 120

90 90

80 w30

-30 FIGURE 7.ELA reconstructions for all glaciers plotted separately 1800 1850 1900 1950 2000 1800 1850 1900 1950 2000 Year Year for six regions.

580 / ARCTIC, ANTARCTIC, AND ALPINE RESEARCH 60 2 40

20 E g0 W FIGURE 8.(A) All ELA recon- C-20. structions (gray) an the average ELA reconstruction (Mack). The -40 o ELA reconstruction of each glacier -60 is shifted compared to Figure 7 in order to get a mean E'(t) of zero -80 over the period 1910 to 1959. (B) Same as Figure 8A, but for tem- 18501870 1890 19101930 1950 1970 1850 1870 189019101930 1950 1970 Year Year perature reconstructions.

The ELAs of Fr(as and San Quint(n show hardly any correspon- We estimated for each glacier the sensitivity of the mass balance to dence. The ELA of San Quintin decreases exceptionally after 1920. As a change in temperature (T'(t)) by using a parameterization of San Quintin is a piedmont glacier and not a valley glacier, it could be Oerlemans (2001): argued whether its ELA reconstruction is reliable. However, the rapid B (t)= -0.271 (Pan")o.51n decrease could be explained by an increase in precipitation between = P (7) 1919 and 1935 (Winchester and Harrison, 1996). Winchester and T,(t) Harrison (1996) also found that precipitation is a more dominant factor Oerlemans (2001) based this parameterization on calculations with than air temperature in controlling the fluctuations of San Quintin. a mass-balance model applied to a set of 13 glaciers. The sensitivity is The ELA of the European glaciers increased between 1920 and a function of mean annual precipitation (Pa,,,,; in in w.e.) and increases 1950. Sorbreen reached a maximum ELA before 1950 and the other for wetter climates. We chose, for each glacier, a value for the annual glaciers after 1950. Morteratschgletscher's increase in the ELA is large precipitation, based on the compiled data of Zuo and Oerlemans compared to the other ELA reconstructions in the European Alps. (1997b) (Table1).Figure 8B shows the calculated temperature Bogen et at. (1989) found that the increase in the ELA of Nigardshreen reconstructions for each glacier and an average temperature trend. that occurred between 1925 and 1955 is related to a series of excessive Between 1910 and 1959, the average increase in air temperature was warm summers. 0.8 K and the standard error in the mean 0.2 K. The calculated linear Marukskij, Bezengi, and Gergeti all show slowly increasing ELAs trend in temperature was 0.19 K per decade. until 1950-1970 and a decrease afterwards. The ELA of Soliyskiy increases rapidly around 1900 and starts decreasing already after 1930. The reconstructed ELAs of Fox and Franz Josef show very similar Discussion patterns, with a rapid increase in the ELA after 1920 until 1960. The high ELAs around 1960 are strongly linked with changes in the Our results agree with those of the IPCC (2001) on an increase in the circulation patterns over the southwest Pacific region (Fitzharris et al., global surface air temperature between 19 10 and 1945, if we assume that 1992). At that time, summer pressures over New Zealand were higher the ELA fluctuations were solely explained by changes in temperature. than normal due to a pole-ward shift of the subtropical high, favoring In addition, cooling during the period 1946 to 1975 in the Northern clearer skies and more ablation. Moreover, snow accumulation was Hemisphere (IPCC, 2001) is consistent with the observed lowering in unusually low in the 1960s and 1970s. our reconstructed ELAs. However, the warming rate reported by the IPCC for the period 1910-1945 was less (0.14 K per decade) than estimated from the ELA reconstructions for 1910 to 1959 (0.19 K per WORLDWIDE PATTERN IN ELA FLUCTUATIONS decade). This may be associated with the fact that we excluded For all glaciers, ELA reconstructions were calculated for the precipitation and cloudiness as possible causes of the ELA fluctuations. period between 1910 and 1959. We therefore estimated an average Besides, we estimated this global temperature increase from only a few worldwide change in the ELA for this period (Fig. 8A). The average glacier records that are also not evenly distributed over the globe. ELA increase was 33 m between 1910 and 1959, with a standard error The IPCC (2001) also concluded that the Southern Hemisphere of 8 m. According to Figure 7, this increase was most pronounced in has been warming more uniformly during the 20th century compared the U.S.A. and New Zealand and less in Asia. Generally, the re- with the Northern Hemisphere and also shows wanning between 1946 constructed ELAs increased strongly during the first 50 yr of the 20th and 1975. This is in contrast with the decreasing ELAs of the glaciers century. Except for San Quintin, they all decreased again in the second in New Zealand, unless precipitation has increased simultaneously. half of the. 20th centuryuntil1980, when most of our ELA Regional differences from the hemispherical mean could also account reconstructions end. This implies that the first half of the 20th century for this discrepancy. was warmer or drier than the period before 1900. After 1960, the Our results do not give rise to the conclusion that there was climate returned to- a situation that favors lower ELAs. a second period (1976 to 2000) of global warming, as claimed by the IPCC (2001). This is because most of our reconstructed ELAs do not extend beyond 1980, due to the time span of the length records and the RECONSTRUCTIONS OF TEMPERATURE Gaussian filter that we applied. If we assume that the changes in the ELA were solely due to As we related changes in the ELA to changes in the specific mass temperature variations, we can derive historic trends in air temperature balance, we could as well determine the mean specific mass balance for from the ELA reconstructions. We then only need to know the sen- each glacier between 1910 and 1959. Averaged over all glaciers, this sitivity of the glacier mass balance to the temperature. A change in the resulted in a mean specific mass balance of -0.18 in w.e. a -l with ELA (E'(t)) can be easily translated into a mass-balance change (B'(t)) a standard error of 0.04 m w.e. a-r. If we assume this value to be by multiplying it with the mass-balance gradient, given in Table 1. representative for all glaciers in the world, the mean rise in sea level

E. J. KLOK AND J. OERLEMANS / 581 due to the retreating glaciers can be roughly estimated. Assuming an Armstrong, R. L.,1989: Mass balance history of Blue Glacier, area of 527 900 km2 for all of the world's small ice caps and glaciers Washington, USA. In Oerlemans, J. (ed.), Glacier Fluctuations and excluding Greenland and Antarctica (Zuo and Oerlemans, 1997b), we Climatic Change. Dordrecht: Kluwer Academic Publishers, 183-192. derived a total rise in sea level of 1.3 cm over the period 1910-1959 Bogen, J., Wold, B., and Ostrem, G., 1989: Historic glacier variations and a mean rate of about 0.3 mm a 1. This rate is in accordance with in Scandinavia. In Oerlemans, J. (ed.), Glacier Fluctuations and the estimated contribution of glaciers to sea-level rise for the period Climatic Change. Dordrecht: Kluwer Academic Publishers, 109- 1910-1990 reported by the IPCC (2001): 0.2 to 0.4 mm a-'. However, 128. Chinn, T. J., 1996: New Zealand glacier responses to climate change of we should realize that this estimate, and likewise the estimate for the the past century. New Zealand Journal of Geology and Geophysics, global temperature increase, is not very accurate, as we did not include 39: 415-428. any information from glaciers in Alaska and northeast Canada, which Conway, H., Rasmussen, L. A., and Marshall, H. P., 1999: Annual are highly glacierized areas. Neither did we use a weighted average to mass balance of Blue glacier,U.S.A.:1955-97. Geografiska account for the size of each glacierized region. Besides, the accuracy of Annaler, 81A: 509-520. the estimates of course depends on the data quality, and the errors in De Smedt, B. and Pattyn, F., 2003: Numerical modelling of historical the model results. front variations and dynamic response of Sofiyskiy Glacier, Altai Mountains, Russia. Annals of Glaciology, 37: 143-149. Dyurgerov, M. B., 2002: Glacier mass balance and regime: data of Summary measurements and analysis. Meier, M. F. and Armstrong, R. (eds.), Institute of Arctic and Alpine Research, University of Colorado, The purpose of this study was to extract information on the past Occasional Paper, 55. 88 pp. climate from glacier length fluctuations by means of deriving historical Fitzharris, B. B., Hay, J. E., and Jones, P. D., 1992: Behaviour of New ELAs. We did this for 19 glaciers from different parts of the world. We Zealand glaciers and atmospheric circulation changes over the past used a simple model that takes into account the geometry of the glacier, 130 years. The Holocene, 2: 97-106. the length response time, and the mass balance-surface height Haeberli, W. and Hoelzle, M., 1995: Application of inventory data for feedback (Klok and Oerlemans, 2003). Because we did not always estimating characteristics of and regional climate-change effects on know the input parameters of each glacier accurately, we carried out mountain glaciers: a pilot study with the European Alps. Annals of Glaciology, 21: 206-212. a sensitivity test for a fictitious glacier. We found that the uncertainty in Hoelzle, M., Haeberli, W., Dischl, M., and Peschke, W., 2003: Secular the characteristic width of the glacier snout, the melt rate at the glacier glacier mass balances derived from cumulative glacier length terminus and the mass-balance gradient had the largest impact on the changes. Global and Planetary Change, 790: 1-12. ELA reconstruction. More specifically, for Nigardsbreen, the standard Hooker, B. J., and Fitzharris, B. B., 1999: The correlation between deviation in the ELA reconstruction due to uncertainties in the input climatic parameters and the retreat and advance of Franz Josef parameters was estimated to be 4 m. Glacier, New Zealand. Global and Planetary Change, 22: 39-48. The results show that all glaciers of our data set experienced an Huybrechts, P., de Nooze, P., and Decleir, H., 1989: Numerical increase in the ELA between 1900 and 1960. The average increase modelling of Glacier d'Argentiere and its historic front variations. In between 1910 and 1959 was 33 m, with a standard error of 8 m. After Oerlemans, J.(ed.), Glacier Fluctuations and Climatic Change. around 1960 until 1980, the ELAs decreased to lower elevations. This Dordrecht: Kluwer Academic Publishers, 373-389. IPCC, 2001: The Scientific Basis. Contribution of Working Group 1 to implies that during the first half of the 20th century, climate was warmer the Third Assessment Report of the Intergovernmental Panel on or drier than before. From the ELA reconstructions, we estimated an Climate Change. Cambridge: Cambridge University Press. 881 pp. average temperature increase of 0.8 ± 0.2 K and a sea-level rise of about Kaser, G., 1999: A review of the modem fluctuations of tropical 0.3 mm a-'. These results support the evidence that an average global glaciers. Global and Planetary Change, 22: 93-103. temperature increase took place between 1910 and 1945 and a sea-level Kick, E. J. and Oerlemans, J., 2003: Deriving historical equilibrium- increase of 0.2 to 0.4 mm a-' (IPCC, 2001). However, care should be linealtitudes from a glacier length record by linear inverse taken when interpreting the global estimates since we did not include modelling. The Holocene, 13: 343-351. length records of glaciers in the highly glacierized areas Alaska and Krimmel, R. M., 1989: Mass balance and volume of South Cascade northeast Canada, neither did we use a weighting function to account for Glacier, Washington 1958-1985. In Oerlemans, J.(ed.), Glacier the size of each glacierized region. The modeled ELA reconstructions Fluctuations and Climatic Change, Dordrecht: Kluwer Academic Publishers, 193-206. reveal regional differences. Changes in the ELA are most pronounced Krimmel, R. M., 2000: Water, Ice and Meteorological Measurements for North America, Europe, and New Zealand. at South Cascade Glacier, Washington, 1999 Balance Year. USGS WRI-00-4139. Oerlemans, J., 1994: Quantifying global warming from the retreat of Acknowledgments glaciers. Science, 264: 243-245. We are grateful to a number of people who supplied us with Oerlemans, J., 2001: Glaciers and Climate Change. Rotterdam: A.A. useful data and information: Ricardo Villalba, Mike Demuth, Frank Balkema. 148 pp. Pattyn, Bert De Smedt, Trevor Chinn, Regula Frauenfelder, Andrey Pelfini, M. and Smiraglia, C., 1997: Signals of 20th-century warming Glazovsky, and Al Rasmussen. We thank the members of the Ice & from the glaciers in the Central Italian Alps. Annals of Glaciology, Climate group of the IMAU and Angelina Souren for their helpful 24: 350-354. comments. We also thank Chris Waythomas and an anonymous Pattyn, F., De Smedt, B., De Brabander, S., Van Huele, W., Agatova, reviewer for commenting on our manuscript. A., Mistrukov, A., and Decleir, H., 2003: Ice dynamics and basal properties of Sofiyskiy Glacier, Altai Mountains, Russia based on DGPS and radio-echo sounding surveys. Annals of Glaciology, 37: References Cited 286-292. Pelto, M. S. and Hedlund, C., 2001: Terminus behavior and response Anda, A., Orheim, O., and Mangerud, J., 1985: Late Holocene glacier time of North Cascade glaciers, Washington, U.S.A. Journal of variations at Jan Mayen. Polar Research, 3, n.s.: 129-140. Glaciology, 47(158): 497-506. Aniya, M., 1999: Recent glacier variations of the Hielos Patag6nicos, Perez Moreau, R. A., 1945: Resena Botanica sobre los Parques South America, and their contribution to sea-level change. Arctic, Nacionales Nahuel Huapi, Los Alerces y Lanfn. Anales del Museo de Antarctic, and Alpine Research, 31: 165-173. la Patagonia, 1: 253-276.

582 / ARCTIC, ANTARCTIC, AND ALPINE RESEARCH Pohjola, V. E. and Rogers, J. C., 1997: Atmospheric circulation and Villalba, R., Leiva, J. C., Rubulls, S., Suarez, J., and Lenzano, L., 1990: variations in Scandinavian glacier mass balance. Quaternary Re- Climate, tree-ring, and glacial fluctuations in the Rio Frias valley, Rio search, 47: 29-36. Negro, Argentina. Arctic and Alpine Research, 22: 215-232. Rasmussen, L. A. and Conway, H., 2001: Estimating South Cascade Warren, C. R. and Sugden, D. E., 1993: The Patagonian Icefields: A Glacier (Washington, U.S.A.) mass balance from a distant radio- glaciological review. Arctic and Alpine Research, 25: 316-33 1. sonde and comparison with Blue Glacier. Journal of Glaciology, Winchester, V. and Harrison, S., 1996: Recent oscillations of the San 47(159): 579-588. Quintin and San Rafael glaciers, Patagonian Chile. Geografiska Reynolds, J. R. and Young, G. J., 1997: Changes in areal extent, Annaler, 78A: 35-49. elevation and volume of Athabasca Glacier, Alberta, Canada, as Zuo, Z. and Oerlemans, J., 1997a: Numerical modelling of the historic estimated from a series of maps produced between 1919 and 1979. front variation and the future behaviour of the Pasterze glacier, Annals of Glaciology, 24: 60-65. Austria. Annals of Glaciology, 24: 234-241. Savoskul, O. S., 1997: Modem and Little Ice Age glaciers in `humid' Zuo, Z. and Oerlemans, J., 1997b: Contribution of glacier melt to sea- and 'arid' areas of the Tien Shan, Central Asia: two different patterns level rise since AD 1865: a regionally differentiated calculation. of fluctuation. Annals of Glaciology, 24: 142-147. Climate Dynamics, 13: 835-845. Solomina, O. N., 2000: Retreat of mountain glaciers of northern Eurasia since the Little Ice Age maximum. Annals of Glaciology, 31: 26-30. Ms submitted June 2003

1 r

E. J. KLOK AND J. OERLEMANS / 583